I am attempting to dock two proteins and we have a general idea of the binding site on one of the proteins (protein 1) , but are unsure where the binding site of the second protein is (protein 2). We'd like to restrict the dockable area of protein 1 while leaving protein 2 unrestricted. I am trying to use the uniform trans flag. I did some testing/internet searching but I can't seem to understand exactly how its working. Here is the definition:
Uniform random repositioning of the second partner about the first partner within a sphere of the given radius, [R].
1. where is the origin of the sphere? can we change this?
2. What is the criteria/cutoffs for "within"?
May be i'm approaching this in the wrong way, any suggestions would be helpful.
Looking at the code, it seems that the uniform trans flag repositions the second docking partner uniformly within a sphere of given radius (centered about the initial position of the second docking partner). Note that this flag will not change the rotation of the second partner relative to the first, only the translation between the two. Think of drawing a sphere around the initial position of the second partner, then dividing that sphere evenly into cubes and picking one of those cubes at random as the new center for the second partner; that's basically what the code's doing. That said, I don't think this is the best flag to use for your situation. In similar situations (where I have some biochemical data on one partner but not the other) I've used a combination of the randomize and dock_pert flags. The flag "randomize2" will completely randomize the orientation of the second partner relative to the first (the translation is unchanged but the rotation is uniformly random). This basically means that there is some chance of attempting to dock every surface region of the second partner to the first (since there is a complete lack of info on the second partner). The dock_pert flag then handles any minor uncertainties you might have in the binding interface of the first partner. dock_pert T R will translate the second partner about the first by a gaussian distributed distance with standard deviation T and rotate by a a random axis by a gaussian distributed angle R. The rotation part is overkill (since you have already randomized the rotation) but the translation is helpful: rather than sampling from a uniform random sphere, the gaussian distribution will ensure that you sample more closely to your initial guess at the binding interface. dock_pert 3 8 is often used if you have a good indication of the interface, but I've gone up to 5 10 and even 8 15 depending on my uncertainty (again the angle is irrelevant because if you've already passed the randomize flag). Hope this helps!